A 5-year, 6-site study, consisting of 6 interlocking RO1 applications, is proposed to study the treatment of SSR-resistant depression in 400 adolescent subjects. Subjects will be those with DSM-IV MDD, currently in treatment, and still depressed despite at least 4 weeks of treatment at an adequate dosage (20mg) and at least 2 weeks of treatment at a higher dosage (40mg) of either paroxetine or fluoxetine. We focus on these two SSRIs because they are most commonly used drugs in this class, and are the only two for which efficacy has been demonstrated in the treatment of adolescent depression. After an initial assessment, the adolescents will be observed at the higher dosage of SSRI for an additional two weeks, and then reassessed. Those who show no significant response over that time (decrease in CDRS-R <20%) will be tapered from their current regimen and entered into the protocol. These 400 subjects will be assigned to one of four conditions to be delivered over 12 weeks. The rate of clinically acceptable response to treatment (defined as a CGI equal to or greater than 2 and equal to or less than 50% decrease in the CDRS-R) will be compared across the 4 cells in a 2x2 factorial design: (1) switch within SSRI class (those on paroxetine switch to fluoxetine; those on fluoxetine switch to paroxetine); (2) switch to a different class of agent (venlafaxine); (3) switch within SSRI class plus receive cognitive behavior therapy (CBT), and; (4) switch to a different class of agent (venlafaxine) plus CBT. Subjects who show a clinically acceptable response will receive 12 additional weeks of continuation treatment with the same intervention as in the acute phase. Non-responders will be offered 12 weeks of open treatment. All subjects will be followed up for 12 months after the continuation phase, regardless of treatment compliance. We hypothesize that there will be a medication effect (venlafaxine superior to SSRI switch), a CBT effect (CBT + medication superior to medication alone), and that CBT + venlafaxine will be superior to the other 3 cells. in addition, we hypothesize that the rate of relapse and recurrence will be lower in the CBT treated cells. The six sites participating, and the numbers of subjects to be enrolled at each are: Brown (40), Dallas (80), Galveston (80), Oregon (80), UCLA (n=40), the site of this application, and Pittsburgh (n=80), with the latter being the coordinating site for the overall study.